☕🧠 JAVA C1 vs C2: where does the JVM put your COMPILED ...

Return to site

☕🧠 JAVA C1 vs C2: where does the JVM put your COMPILED CODE?

Your Java bytecode does not stay bytecode forever. 🔥

· java

🔸 TLDR

HotSpot can progressively transform frequently executed Java bytecode into native machine code:

▪️ Interpreter → immediate execution 

▪️ C1 → fast compilation 

▪️ C2 → highly optimized compilation 

▪️ Generated native code → stored in the Code Cache

And no: C1/C2 compiled code does NOT live in the Java heap.

🔸 1️⃣ START WITH BYTECODE

static int square(int x) {
    return x * x;
}

public static void main(String[] args) {
    for (int i = 0; i < 10_000_000; i++) {
        square(i);
    }
}

At startup, square() exists as JVM bytecode.

The JVM can initially interpret it.

But after enough executions, HotSpot notices:

"This method is HOT." 🔥

#Java #JVM

🔸 2️⃣ C1: COMPILE FAST

C1 is the Client Compiler.

Its goal is roughly:

▪️ Compile quickly 

▪️ Improve performance early 

▪️ Collect profiling information

▪️ Avoid spending too much CPU optimizing

So the JVM can quickly replace interpreted execution with native instructions.

Java bytecode
      ↓
     C1
      ↓
Native machine code

#JIT #HotSpot

🔸 3️⃣ C2: OPTIMIZE HARDER

If a method remains hot, the JVM may eventually use C2.

C2 spends more compilation time looking for optimizations such as:

▪️ Method inlining 

▪️ Dead-code elimination 

▪️ Escape analysis

▪️ Loop optimizations

 ▪️ Speculative optimizations

Bytecode
   ↓
Interpreter
   ↓
C1 compiled code
   ↓
C2 highly optimized code

This is one reason why a Java application can become faster after warming up. 🚀

#Performance #JavaPerformance

🔸 4️⃣ BUT WHERE IS THAT CODE STORED? 🧠

Not here:

Java Heap

Java Heap
├── Objects
├── Arrays
└── Application data

The generated machine code goes into native JVM memory called the:

👉 CODE CACHE

Conceptually:

JVM Process Memory
│
├── Java Heap
│   └── Java objects
│
├── Metaspace
│   └── Class metadata
│
├── Thread Stacks
│   └── Frames / local variables
│
└── Code Cache
    └── JIT compiled native code
         ├── C1 code
         └── C2 code

So intensive JIT compilation consumes JVM process memory without necessarily increasing your heap.

#Memory #CodeCache

🔸 5️⃣ WATCH IT HAPPEN

Run:

java -XX:+PrintCompilation Main

You may see output similar to:

35   5       3       Main::square (4 bytes)
 40   6       4       Main::square (4 bytes)

The compilation level is especially interesting.

With tiered compilation, HotSpot commonly uses levels:

0 → interpreted
1 → C1
2 → C1
3 → C1 + profiling
4 → C2

So seeing level 4 means the method reached the JVM's highly optimized compilation tier. 🔥

🔸 TAKEAWAYS

▪️ C1 and C2 are JIT compilers, not memory regions.

▪️ C1 favors fast compilation and profiling.

▪️ C2 spends more effort producing highly optimized native code.

▪️ Hot methods can move progressively through compilation tiers.

▪️ JIT-generated machine code is stored in the Code Cache.

▪️ The Code Cache belongs to JVM native/process memory, not the Java heap.

The JVM does not simply "run Java code." It watches your application, learns what is hot, and recompiles it accordingly. ☕🔥

#java #jvm #jit #hotspot #c1 #c2 #javaperformance #performance #memory #codecache #java27 #javadeveloper #softwareengineering #javacertification #ocpjava

🔸 SOURCES & FURTHER READING 📚

Want to verify how C1, C2, Tiered Compilation and the Code Cache really work inside HotSpot? ☕⚙️

▪️ Oracle: Compilation Optimization: C1 & C2 JIT Compilers https://docs.oracle.com/javacomponents/jrockit-hotspot/migration-guide/comp-opt.htm

▪️ Oracle: HotSpot Performance Enhancements: Tiered Compilation & Code Cache https://docs.oracle.com/en/java/javase/24/vm/java-hotspot-virtual-machine-performance-enhancements.html

Straight from the JVM documentation. 🔎☕

#Java #JVM #HotSpot #JIT #C1 #C2 #JavaPerformance #CodeCache

Go further with Java certification:

Java👇

https://bit.ly/javaOCP

Spring👇

https://bit.ly/2v7222

SpringBook👇

https://bit.ly/springtify

JavaFullstackBook👇

https://bit.ly/jfstack27